Back to portfolio
Voice / Multimodal · Finance

Becon

Voice Financial Coach

Financial forecasting dashboard with integrated Eleven Labs voice AI coach — lets business operators get real-time cash flow insights and have voice-to-voice conversations about their financial data.

01The Problem

Business operators need to understand their cash position, forecast runway, and financial KPIs — but most financial dashboards are static reporting tools that answer questions the operator already thought to ask. They don't proactively surface insights, and they require reading and interpreting charts. A voice AI layer on top of financial data changes the interaction model: instead of "open dashboard, find the right chart, interpret it," you ask "how's my runway?" and get an answer in spoken language with context.

02What the AI Does

A Fathom HQ-style financial dashboard showing cash flow visualization, interactive charts, and KPI tracking. Integrated into this dashboard: an Eleven Labs Conversational AI voice coach that operators can speak to directly. The voice coach has access to the financial data and can answer questions about cash position, runway, burn rate, and forecast scenarios in real-time voice conversations. High-quality 16kHz audio streaming with echo cancellation. Built on: Node.js + Vite frontend, Eleven Labs Conversational AI (voice), Fathom-style chart visualizations.

03Design Decisions

01 · Choice

Voice as the primary query interface for financial data

Why

Reading charts requires visual attention and interpretation. Spoken answers require nothing but listening. For a busy business operator, "what's my burn rate?" answered in 10 seconds by voice beats navigating a dashboard.

Constraint

Voice queries require the financial data to be structured enough for the AI to answer accurately. If the underlying data model is weak, the voice layer will produce confident errors.

02 · Choice

Eleven Labs Conversational AI as the voice engine

Why

Eleven Labs has one of the strongest voice AI platforms for real-time conversation. The conversational AI agent format maps well to the "ask a question, get a spoken answer" use case. **[Creator: add rationale]** for why Eleven Labs was chosen over alternatives (Vapi, Pipecat, etc.) and whether the voice layer is connected to live financial data or a static dataset.

03 · Choice

Fathom-style visualization as the baseline UI

Why

Fathom (a financial planning tool) has a well-regarded visual language for cash flow charts. Borrowing that visual paradigm reduces the "what am I looking at?" friction for users already familiar with modern financial dashboards.

Constraint

The visual similarity to Fathom is a starting point, not a destination. As the product evolves, the UI should develop its own visual identity.

04Tradeoffs & Limits

- **Voice layer accuracy is constrained by data model quality.** If the underlying financial data isn't structured or updated in real-time, the voice AI will answer questions with stale information and not know it. - **Financial data is sensitive.** A voice-accessible financial assistant has higher confidentiality stakes than a general chatbot. Any voice recording or transcript storage has GDPR and business data handling implications. - **Echo cancellation and audio quality vary by device/microphone.** The "high-quality 16kHz audio" experience is best-case; real-world usage on laptop microphones or in noisy environments will be degraded. - **No autonomous alerts or anomaly detection.** The voice layer responds to queries but doesn't proactively surface problems (e.g., "your burn rate just spiked"). Proactive alerting requires a separate monitoring layer.

05Key Insight

Voice interfaces on top of structured data are most valuable when the alternative is "open a dashboard and figure it out." The efficiency gain isn't in getting answers faster — it's in eliminating the navigation and interpretation overhead that causes people to not ask questions in the first place.